Feature Business Continuity
A Software Bug at One Bank Read as an Outage at Four. Check Your Dependency Map.
The Nov. 16 cross-bank payment outage exposed the blast radius in shared rails. Build the BIA dependency row your business continuity plan is missing.
Table of Contents
TL;DR
- On November 16, 2026, Downdetector logged 4,810 reports against Chase, 2,045 against Bank of America, 1,072 against U.S. Bank, and 1,059 against Wells Fargo — over a software bug that originated at exactly one of those four firms (Cyber Magazine).
- Bank of America and U.S. Bank both said their own systems were fine. Their customers disagreed, because the failure surfaced at the other end of cross-institution Zelle payments.
- A BCP that maps vendors but not counterparty rails has no row for this event. Build the row before you need it: rail, counterparty path, MTD by payment type, manual fallback, external signal, owner.
Pick the number that matters: 4,810. That is how many people reported Chase as broken in a fifteen-minute window on a Monday afternoon — the bank’s second-highest reading in at least a year, behind only September 24 (USA TODAY).
Now the number that should worry you more: 1,072. That is U.S. Bank’s peak, and 1,059 is Wells Fargo’s. Both firms said their own systems were not the problem. Both were correct, and both still had customers staring at an app that would not move money.
That gap — accurate statement, unhappy customer — is the whole continuity problem in one frame. Your dependency map either has a row for it or it does not.
What actually happened
The sequence is short and worth reading closely, because the shape of a failure tells you what controls were missing.
Zelle complaints started climbing around 9:30 a.m. ET and peaked near 2 p.m. Chase’s Downdetector volume spiked in the 2:00–2:15 p.m. window. At 2:10 p.m. a Chase customer in Texas was told by a Dallas branch teller that “the system was down,” with no further detail (CBS News). By roughly 7 p.m. Chase said the issue was resolved and transactions were processing normally.
Chase’s explanation, posted to X:
“JPMorganChase has experienced a temporary issue related to a software bug in on-premise hardware systems – there is no evidence this is cyber-related.”
And on scope:
“Where we were party to a transaction involving other banks, those banks may have been impacted, but this is not a systemic issue.”
The propagation path ran through Zelle, which is operated by Early Warning Services — a company owned by Bank of America, Capital One, JPMorgan Chase, PNC, Truist, U.S. Bank, and Wells Fargo. Zelle moved $1.2 trillion across 4.2 billion transactions in 2025 (Early Warning Services). A dispute over whether the rail or the bank was broken is beside the point: the network is owned by the same seven firms whose names appear in the Downdetector charts.
Two failures compounded. A software defect in on-premise hardware, and a shared rail that turned one firm’s defect into seven brands’ customer experience.
”Not directly impacted” is a status customers do not recognize
Bank of America said its “systems have been operating normally today.” U.S. Bank said its systems were not directly impacted while warning that Zelle payments with another institution might be delayed. Both statements were defensible.
Neither stopped the phones.
Mike Chapple, an IT and analytics professor at Notre Dame, described the mechanism to CBS: the outage “appears to stem from a technical issue at Chase that affected other banks through their shared payment networks.” Karen Panetta, dean of graduate engineering at Tufts, put the consequence plainly — backbone providers have reach, and “when one of those backbone providers has disruptions, the downstream effects are magnified.”
If your BCP’s scope statement defines an incident as something inside your perimeter, your plan cannot describe November 16. The FFIEC Business Continuity Management booklet already expects more: section III.A.2 is interdependency analysis, and it does not distinguish between a core vendor and a counterparty bank you settle through.
The BIA row you are missing
Most business impact analyses are organized by process and then by vendor. Counterparty rails get buried in prose somewhere in the payment section, if they appear at all. The core-provider concentration problem is usually where teams first notice the gap, and this event is the same shape with a different name on the door.
Pull the row out and make it structural. One row per rail, per path:
| Field | Example entry |
|---|---|
| Rail and path | Zelle, outbound to counterparty bank via EWS directory |
| Process supported | Retail P2P, small-business vendor payments |
| MTD for this payment type | 4 hours (consumer), next settlement window (scheduled vendor) |
| Manual fallback | ACH same-day batch; wire for urgent; branch cash |
| External signal | Downdetector brand velocity; counterparty status page |
| Owner | Director of Payment Operations |
| Last tested | Tabletop, 2026-Q3, result linked |
Fill the MTD by payment type, not by platform. “Zelle: 4 hours” is meaningless; “rent payment, 4 hours” and “contractor payroll, next window” are decisions someone can execute at 2 p.m. on a Monday.
Section V.D of the same booklet expects alternate arrangements and manual call-in procedures for correspondent institutions to sit in the plan document, not in an appendix someone remembers exists. And section IV.A.8 covers change management as a resilience domain — relevant here, because the trigger was a software change to on-premise hardware.
The backlog half of this is covered in detail in the Fedwire/FedNow outage piece: a written order of battle for the queue, agreed before the queue exists. Same rule applies to a counterparty rail. Decide what goes first at 1:55 p.m., while you still have time to think.
Detection you do not own
When the failure is at a counterparty, your synthetic monitoring can be green while customers are red. You need a feed that watches the outside world.
Downdetector is imperfect and self-selected, but it is fast and it is directional. Treat it as a monitored source with a numeric escalation trigger, written down and version-controlled.
Starter threshold, to calibrate against your own trailing three to six months:
- Amber: report velocity at three times the trailing 15-minute baseline for your brand, sustained across two consecutive readings.
- Red: amber plus contact-center queue abandonment over 15 percent, or any confirmed counterparty status-page entry.
Anti-gaming check: pull the corresponding call-driver codes the same day and confirm the Downdetector spike maps to actual Zelle or transfer contacts. A spike without call volume is noise; a spike with it is your afternoon.
Assign the read to a named person on each shift, with a timestamped log. An escalation nobody can audit is an anecdote.
The first ninety minutes, in order
- 0–10 min. Open the payment ops bridge. Confirm the rail path, not the symptom. Log the first Downdetector reading and the counterparty status page state.
- 10–25 min. Determine whether the failure is inbound, outbound, or both. Originating and receiving failures need different customer messaging.
- 25–40 min. Freeze duplicate resubmissions. Customers who re-tap a stalled Zelle send at 2:05 p.m. are creating a reconciliation problem at 4 p.m.
- 40–60 min. Publish one internal status line and one customer line. Name the rail, name the counterparty generically, give the next update time. Chase’s own customer-facing copy — “some services aren’t available… please try back later” — is a reasonable minimum.
- 60–90 min. Stand up the exception queue. Every failed transaction gets an ID, a timestamp, and an owner before anyone starts fixing balances.
The morning after
Reconciliation is where cross-institution events bite. Duplicate sends, sends marked failed that actually settled, and received funds that posted twice all surface after the apps go green.
Run three queries before the branch opens: transactions in a pending state longer than four hours, any transaction with two settlement records, and every customer contact tagged to the rail in the prior 24 hours. Screenshot evidence is already in the file if your contact center captured it — the advice experts gave customers during this outage, to screenshot every error, was also free evidence for the bank (CBS News).
Large firms have a written standard here. SR 20-24, the interagency sound practices paper, was revised June 2, 2026 to strip reputational-risk references and still applies at $250 billion in assets, or $100 billion with $75 billion in any one concentration measure. It draws its practices from business continuity, third-party, and cyber risk management and is directed at the largest, most complex firms. If you are under the threshold, it is still the cleanest list of what good looks like.
So what
The failure lasted about five hours and produced no cyber finding. It also produced roughly nine thousand public complaints across four brands, a branch teller improvising, and a shared rail that carried one firm’s defect into six other banks’ customer bases.
Your plan gets one of two rows this week. Either a counterparty-rail dependency row with an MTD, a fallback, an owner, and an external signal — or a gap you find out about the next time someone else’s release goes wrong.
The BCP/DR Kit ships the BIA workbook with dependency and MTD fields already built, plus the exercise templates for the tabletop above.
FAQ
Does a counterparty outage count as ours? For continuity, yes. The customer does not distinguish. Build the row and own the experience of a failure you cannot fix.
What belongs in the fallback field? A method your staff can execute without a project. Same-day ACH batch, wire for urgent value, branch cash, a stated hold-and-release rule. “Escalate to vendor” is not a fallback.
How often should the row be tested? Run the tabletop annually and after any counterparty change. The FFIEC booklet’s section VII.I and VII.J set the third-party and core-firm expectations; the vendor-BCP integration walkthrough covers how those expectations get examined. Run the internal session first — it tells you which handoffs break before a counterparty is in the room.
◆ Need the working template?
Start with the source guide.
These answer-first guides summarize the required fields, evidence, and implementation steps behind the templates practitioners search for.
◆ Related template
Business Continuity & Disaster Recovery (BCP/DR) Kit
Set recovery priorities from a business impact analysis, catch vendors that recover slower than the processes they support, run and log tabletop exercises and tests, and report to the board on one page. Excel workbook, guide, tabletop kit, case studies and Word templates.
◆ Immaterial Findings · Weekly
Sharp risk & compliance insights. No fluff.
◆ FAQ
Frequently asked questions.
How do you write a dependency row for a payment rail in a BIA?
Is a counterparty bank's outage our incident if our own systems are fine?
What is a reasonable external-signal threshold before escalating?
Should we test a counterparty outage scenario with the counterparty?
Author
Rebecca Leung
Rebecca Leung has 8+ years of risk and compliance experience across first and second line roles at commercial banks, asset managers, and fintechs. Former management consultant advising financial institutions on risk strategy. Founder of RiskTemplates.
◆ Related framework
Business Continuity & Disaster Recovery (BCP/DR) Kit
Set recovery priorities from a business impact analysis, catch vendors that recover slower than the processes they support, run and log tabletop exercises and tests, and report to the board on one page. Excel workbook, guide, tabletop kit, case studies and Word templates.
◆ Keep reading
Related posts.
Business Continuity
When Fedwire or FedNow Is Down: A BCP That Assumes the Rail Returns
Fedwire and FedNow outage business continuity: settlement windows, a backlog triage rule, Operating Circular 6 Appendix B, and a BIA addendum.
Nov 15, 2026
Business Continuity
BCP Change Management: The Product and Vendor Changes That Should Trigger a Mid-Cycle Continuity Review
Your annual BCP review passed board approval in January. Since then you've migrated to a new core system, added a critical vendor, and launched a new product. None of those changes triggered a BCP update. The FFIEC BCM booklet lists changes like those as triggers for updating the program.
Oct 7, 2026
Business Continuity
The FFIEC CAT Is Retired. The OCC Just Updated Its Exam Checklist to NIST CSF 2.0. Here's the Govern Function Gap Most Financial Institutions Are Missing.
OCC Bulletin 2026-48 aligned the OCC's Cybersecurity Supervision Work Program to NIST CSF 2.0 — including the new Govern function that didn't exist when most financial institutions built their cybersecurity programs. Here's what examiners are now checking and where programs are falling short.
Oct 3, 2026